As in Edit Clusters, the Helm panel is divided into spectra on the right and textual information on the left. First, consider the left-hand portion. The information displayed pertains to a particular cluster, designated in the “Cluster” window, and also specified by the small box in the “Fragments” window. You can scroll through cluster data using the up and down triangle buttons, and you can also click on a cluster number in the Fragments display to call up its information. The “Sequence” panel at the bottom shows your protein’s sequence. It is uninformative in this figure, but as fragments can be associated, then assigned, to particular stretches of residues, the letters will change color. Similarly, the “Match Scores” window is relevant only after the first link has been made, so we’ll discuss that shortly, but not now.
The “Matching Clusters” panel bears further examination, so we’ll examine it in an expanded figure:
The goal in the Edit Links mode is to establish links between clusters. Each cluster, identified in the “Cluster” window, is paired up with a number of other clusters in each direction of the backbone; candidates for the N-terminal link are presented in the left-hand panel, and those for the C-terminal link are in the right-hand panel. Your job is to evaluate which of these clusters in each list is the best, highlight it, and click “Confirm” to establish a link. To undo a link, click “Unconfirm”, and to switch to focusing on one of the peaks in the panel’s list, highlight it and click “GoTo”.
Each list of clusters presents a number of parameters that will help you decide which is the most appropriate link. The left-hand number is the cluster’s number. The next number over, given to three significant figures, is a score indicating the probability that this cluster matches the on under consideration. Displayed scores range from 0.20 to 3.0, and embody how many peaks match in each cluster, their signs (± and how similar their chemical shifts are. Clusters are sorted in the panel according to their match score. Clusters with scores over 2.20 are usually good choices. The middle column, showing integers, indicates how many pairs of the peaks in the two clusters share chemical shifts. The next column indicates sort of a reciprocal preference: the label of the cluster that best matches the candidate cluster. For instance, cluster 0’s best N-terminal match is #9, with a score of 2.20. Cluster 9’s best C-terminal match is #0, so the #9-#0 link is probably good. Custer #40’s best C-terminal match is also #0, though the pair exhibit a low match score; this is probably a poor match, and cluster #40 will need to be matched with a cluster that is not #0. The next column indicates three characteristics of the potential match: R, “Reciprocal Matching”, i.e., each cluster gives the other a high probability of matching; A, “Available”, i.e., not linked in a cluster yet; V, “Viable”, i.e., three or more carbon frequencies match one another across spectra, and the proposed pair exhibits chemical shifts consistent with a pair of residues in the protein sequence.
By default, the top ten best cluster matches are shown in the list. You can change the number that will be displayed by setting the "Max Clusters" parameter in the Parameters tab.
Clicking on a cluster in the list of clusters will update the appropriate peaks in the peak display. Here’s what it looks like when cluster 0 is being examined and clusters 9 and 77 are being considered as link partners. Recall that peaks are labeled with their cluster number preceded by the letter “c”.
Figure 11.18. Editing Links - Cluster #0 matched up with clusters #9 and #77 to the N- and C-terminal sides, respectively
If you switch to examining cluster 20 as a potential N-terminal link mate, you’ll see the left-hand column of spectra change:
Figure 11.19. Editing Links - Cluster #0 matched up with clusters #20 and #77 to the N- and C-terminal sides, respectively
Here’s a figure that should help you figure out which peaks should be lining up with which:
If you think you’ve found a good match, click the Confirm button. You should see something like this if you highlighted N-terminal cluster 9 and clicked the left-hand/N-terminal Confirm button:
Several things happen when you click Confirm. 1) The panel becomes highlighted in yellow. 2) In the Fragments display, the number of the cluster you’re linking to changes position so it is adjacent to your current cluster; here, clicking to link cluster 0 to cluster 9 moves the numeral 9 immediately above the numeral 0 and highlights the area between them in yellow. 3) New information suggesting potential residue assignments appears in the “Match Scores” section of the Helm panel. 4) The “A” becomes an “_” to indicate it’s no longer available.
The RunAbout software examines your clusters and their links, and compares them to those expected for your protein and continuously tries to match your linked fragments to stretches of amino acids. The best potential matches are listed in the Match Scores panel. Column 1 in this display shows the number of N-terminal residue in the proposed matching segment; the final column shows the amino acids in the proposed fragment. In the figure shown here, the confirmed link between clusters 9 and 0 is proposed to correspond to the three residue segment “IFA” beginning with residue I44. Column 2 shows whether this sequence fragment has been assigned already; “1” indicates it has not, and “0” indicates it has. Column 3 shows a Bayesian score (from 0 to 1) that assesses the probability that the assigned clusters match the proposed amino acid segment. The high this number is, the better, when comparing different segment choices, but its value necessarily goes down as the fragments get longer.
If you now click the right-hand/C-terminal Confirm button to link cluster 0 to cluster 77, you will observe similar changes:
When assessing the possible assignment of a cluster it is useful to see not only the experimental data involving the CB,CA an C atoms but additional sidechain atoms as measured in TOCSY experiments. At any cluster in either the Cluster Edit or Link Edit modes you can pop up a Tocsy-CO-NH display. Do use this feature you need to specify CCcoNNH Tocsy and/or HCcoNNH Tocsy peak lists in in the "Parameters" tab. Clicking the "T" button in the Helm tab will then display a strip for these datasets at the HN-N position of the current cluster. These will be displayed in a separate window from the main RunAbout window. The display will be updated as you move to new clusters. At the left side of the display that appears is a list of the 20 amino-acids. Move the mouse over each entry to see a label on the spectra displayed at mean-chemical shift of the carbon and hydrogen atoms of that amino acid type.
The scoring of fragments is based on the joint probability of the assignments of clusters in a fragment. Each cluster gets a score, based on the match of chemical shifts for available peaks.
For each peak in cluster: The score is the ((ppmPeak-ppmAtom)/sdev)^2 where ppmPeak is the chemical shift of the peak ppmAtom is the chemical shift expected for the atom type (from BMRB) sdev is the standard deviation for the atom type (from BMRB)
This gives a number from 0 to some large number, where 0 is for a perfect match.
The scores for all peaks in cluster are summed and a probability is calculated from a Chi square distribution based on the number of peaks, and the total score.
This number runs from 1.0 for a perfect match, to 0.0, for the worst possible match.
If probability is less than the "ppm limit" value (which I realize is a bad name) for any cluster in the fragment the whole fragment is rejected.
The probabilities for each cluster in a fragment are multiplied together to get the final score. Because of this, longer fragments have lower scores. The final score shown is really only useful for comparing possible assignments of a single fragment, rather than different fragments (especially of different lengths).